Join Date: Sep 2006
Attentional control in PD is dependent on genotype and dopamine levels
I'm posting large sections of the article because I don't think it's readily accessible online. I have university access. It's all so complicated. No wonder different neuros have different ideas.
Brain Advance Access published online on January 4, 2008
Attentional control in Parkinson's disease is dependent on COMT val158met genotype
Caroline H. Williams-Gray1, Adam Hampshire2, Roger A. Barker1,* and Adrian M. Owen2,*
Cognitive deficits occur even in the earliest stages of Parkinson's disease. Some such deficits are known to relate to dysfunction in dopaminergic frontostriatal networks, and may be influenced by a common functional polymorphism (val158met) within the catechol O-methyltransferase (COMT) gene. Abnormal attentional shifting behaviour is an important and well-recognized cognitive problem in PD, but nonetheless its precise cognitive and neural basis remains unclear. Here we explored this impairment in an fMRI study employing a recently developed cognitive task designed to fractionate components of attentional control. We investigated the impact of the COMT val158met genotype and dopaminergic medication on both patterns of behaviour and associated brain activation in 29 medicated patients with early PD. Genotype had a critical impact on task strategy: whilst patients with high activity COMT genotypes (val/val) adopted a typical approach of preferentially shifting attention within rather than between dimensions, those with low activity genotypes (met/met) failed to adopt such a strategy, suggesting an inability to form an attentional ‘set’. Moreover, this behaviour was associated with significant underactivation across the frontoparietal attentional network. Furthermore, we demonstrated an interactive effect of COMT genotype and dopaminergic medication on task performance and BOLD response. Hence we have shown for the first time that attentional control in PD is critically determined by genetic and pharmacological influences on dopaminergic activity in frontoparietal networks. This has important implications for understanding the neurobiological basis of attentional control, and highlights the risk of medication-induced cognitive dysfunction in certain genotypic groups of PD patients, which may ultimately impact on clinical practice.
Parkinson's disease (PD) is a common neurodegenerative disease which is defined clinically in terms of bradykinesia, rigidity, tremor and postural instability. However, cognitive symptoms also form an important part of the syndrome, with executive dysfunction being particularly prominent in early disease (Foltynie et al., 2004aGo; Muslimovic et al., 2005Go). The dysexecutive syndrome encompasses difficulty in planning, organizing and regulating goal-directed behaviour, similar to that seen in patients with frontal lesions, and is demonstrable on tasks of attentional control, planning and working memory (Owen et al., 1992Go, 1995Go). The observed influence of dopaminergic medication on performance (Lange et al., 1992Go; Owen et al., 1995Go) together with functional imaging data (Owen et al., 1998Go; Mattay et al., 2002Go), supports the theory that the basis for the dysexecutive syndrome in PD lies within dopaminergic frontostriatal networks. However, whilst loss of nigrostriatal dopamine occurs by definition in all PD patients, not all patients exhibit neuropsychological evidence of dysfunction in downstream frontostriatal networks (Lewis et al., 2003aGo). Furthermore, the relationship between dopamine and executive function is complex, with levodopa administration being associated with an improvement in performance on certain executive tasks, and a deterioration in others (reviewed in Cools, 2006Go). Thus, the precise cognitive and neural bases for executive dysfunction in PD and its heterogeneity are not fully understood.
Some insight into this problem can be gained through exploiting the existence of a common functional polymorphism in catechol O-methyl transferase (COMT), an enzyme which plays a key role in regulating dopamine levels in cortical areas in particular (Karoum et al., 1994Go; Gogos et al., 1998Go; Mazei et al., 2002Go) where, in contrast to the striatum, there are low numbers of dopamine transporters (Lewis et al., 2001Go). The polymorphism results in a substitution of valine for methionine at codon 158 (val158met), which is known to increase the thermostability of the protein and enzyme activity in blood and liver samples (Scanlon et al., 1979Go; Weinshilboum and Dunnette, 1981Go; Boudikova et al., 1990Go; Lotta et al., 1995Go). Furthermore, Chen and colleagues recently measured mRNA expression, protein levels and enzyme activity in post-mortem human prefrontal cortex (PFC) and demonstrated a 40% increase in enzyme activity at 37°C in association with the val variant, mediated principally through altered protein integrity, whilst in vitro work provided further evidence for increased thermostability of the protein as the underlying mechanism for this functional change (Chen et al., 2004Go). We have previously demonstrated that this functional polymorphism has phenotypic consequences in PD. Specifically, low activity COMT genotypes (met/met) leading to higher dopamine levels in the PFC, are associated with impaired performance during the Tower of London planning task (Foltynie et al., 2004bGo) and a reduction in associated frontoparietal activation (Williams-Gray et al., 2007bGo). When interpreted on the background of other studies demonstrating a hyperdopaminergic state in the PFC in early PD compared to healthy controls (Rakshi et al., 1999Go; Kaasinen et al., 2001Go), this work implies that further genetically determined elevations in prefrontal dopamine are detrimental to cognitive performance in PD patients, possibly due to an ‘overload’ effect whereby signal-to-noise ratios (SNR) are reduced (Foltynie et al., 2004bGo). Hence it is unsurprising that the majority of studies investigating the impact of dopaminergic medication on executive performance have produced inconsistent results (Cools, 2006Go), given that this genetic influence has not been considered. Stratifying patients according to COMT genotype not only provides a convenient means of exploring the dopaminergic basis of executive deficits in PD, but is essential in order to interpret medication effects. In this study we have adopted such an approach in order to allow us to explore the neural basis of a key executive deficit in PD, namely the attentional set shifting impairment.
Impaired ability to shift attention between stimuli (i.e. shift attentional ‘set’) is a well-established part of the dysexecutive syndrome in PD and is likely to have important consequences for everyday activities requiring cognitive multitasking. However, this deficit remains poorly understood, in part because tasks traditionally used to examine set shifting ability, such as the Wisconsin card sorting test (Grant and Berg, 1948Go), confound several behavioural components whose neuroanatomical and neurochemical bases may differ. The CANTAB ID/ED shift task, which was developed in an attempt to isolate the set-shifting process (Downes et al., 1989Go), indicates a particular deficit in extradimensional (ED) shifting (between stimulus dimensions) rather than intradimensional (ID) shifting (within stimulus dimensions) in PD (Downes et al., 1989Go; Owen et al., 1992Go, 1993Go). Nonetheless, this shifting deficit still remains poorly defined due to the persistent problem of confounding of multiple cognitive processes at the ED shift stage of this task.
Previous attempts to explore the neurochemical basis of attentional shifting deficits in PD have relied on pharmacological modulation of dopamine levels and have produced inconsistent results (Downes et al., 1989Go; Owen et al., 1992Go; Lange et al., 1992Go; Owen et al., 1993Go; Lewis et al., 2005Go; Slabosz et al., 2006Go). The neuroanatomical basis of the deficit is also uncertain. Whilst functional imaging studies in healthy controls have demonstrated that switching between tasks is mediated by the PFC (Dove et al., 2000Go; Sohn et al., 2000Go; Cools et al., 2002Go), there is a paucity of literature examining the neural basis of the attentional shifting deficit in PD. Furthermore, most previous neuroimaging studies have focused on behaviour induced by externally imposed cues, rather than examining internally generated shifts in attention, which are arguably much more relevant for everyday life, influencing an individual's ability to make decisions and solve problems independently.
Here, we have employed a recently developed cognitive task (Hampshire and Owen, 2006Go), derived from the CANTAB ID/ED task, to examine for the first time internally guided attentional shifting in PD and its neural basis. This task involves working out by a process of trial and error which object is the target in a stimulus set consisting of two faces and two buildings (Fig. 1). The subject makes a series of self-directed ID (e.g. face to face) and ED (e.g. face to building) shifts during this working out period. The task allows the constituent components of attentional control, including responding to novel stimuli, shifting attention both intra- and extra-dimensionally, inhibiting the response to a previously relevant stimulus, and responding to positive and negative feedback, to be examined separately without the confounding which complicates the original CANTAB ID/ED task. Furthermore, the subject's focus of attention is monitored and used to define switching events, rather than attentional switches being imposed externally by the experimental paradigm, thus the problem-solving strategy adopted by the individual can be examined. The task has been validated in healthy controls, in whom individual components of attentional control have been shown to reliably activate neuroanatomically discrete regions of the frontoparietal attentional network (Hampshire and Owen, 2006Go).
Fig. 1 A typical sequence of events during the experimental task. The subject is instructed to work out by a process of trial and error which is the correct target amongst a compound stimulus set comprising two faces superimposed on two buildings. Responses are guided by positive or negative feedback provided after every second response. Once the correct target is identified, the subject continues to respond to the known target, as per prior instruction. After six correct responses (three ‘correct’ feedback messages), either a change of reward contingency occurs, such that the previous target is no longer correct and the subject is required to shift attention to a previous non-target (‘reversal’) or a new stimulus set is presented and the process of working out the correct target begins again (‘set change’).
We directly compared performance, fMRI activation patterns and the impact of dopaminergic medication during this attentional control task in two matched groups of PD patients, differing only in whether they possessed high (val/val) or low (met/met) activity COMT genotypes. Our aim was to define the neural basis of attentional control dysfunction in PD through delineating the impact of genetically and pharmacologically based variation in dopamine levels on behaviour and associated brain activation.
In this study, we have clarified the nature and underlying neural basis of the attentional control deficit which commonly occurs in PD. We have demonstrated for the first time that both genetically determined and pharmacological variation in dopamine levels impact on set-shifting performance and corresponding BOLD activation in a frontoparietal attentional network. Our work confirms that this key executive skill has a dopaminergic basis, but reveals that the influence of dopamine is complex and non-linear.
The observed behavioural effects of COMT genotype on task performance are particularly interesting. We have previously demonstrated a detrimental effect of an increasing number of met alleles (i.e. lower COMT activity) on executive function in early PD using the Tower of London planning task (Foltynie et al., 2004bGo). However, here, rather than observing a universal impairment across behavioural measures in met/met individuals, we identified a difference in the ID/ED response pattern in this group which suggests the adoption of an abnormal problem solving strategy. Typically, controls performing ID/ED shifting tasks preferentially shift attention within rather than between dimensions, and consequently identify the correct target with fewer errors when an ID shift is required rather than an ED shift (Hampshire and Owen, 2006Go). Amongst our PD patients, a similar pattern is seen in the val/val group. However, the met/met group perform equivalently in terms of number of errors whether an ID or ED shift is required, indicating that they do not form an attentional ‘set’ to the previously relevant stimulus dimension, but rather treat each problem independently (Fig. 2). This alternative strategy, although ‘abnormal’, is not actually detrimental in terms of number of errors overall whilst working out the target, and in fact remediates the ED shifting impairment classically observed in PD to some extent. However, in terms of the alternative performance measure of response time, there is a strong trend towards slower performance across all types of event involved in working out the target (ID shift, ED shift, response to set change and response to reversal) in met/met versus val/val patients (Fig. 4). Analysis of the fMRI data suggests that impaired ability to form an ‘attentional set’ and prolonged response times in met homozygotes reflects under-recruitment of the frontoparietal areas necessary for the task (Fig. 6A). Hence our data suggest that impaired attentional control in met versus val homozygotes in PD is due to reduced activity within the PFC, but also reveal that the pattern of attentional shifting impairments in PD is more complex and heterogeneous than the simple ED shifting deficit previously described in the literature.
Reversal learning, i.e. the ability to switch attention from a previous target to a previous non-target following a reversal of reward contingency, was not significantly influenced by COMT genotype in this study. A previous study in healthy controls administered the mixed dopamine and noradrenaline agonist methylphenidate produced analogous results: methylphenidate administration resulted in fewer ED shift errors associated with impaired speed of response, as in our met/met group, but did not influence errors during reversal learning (Rogers et al., 1999Go). These findings, together with evidence from lesion studies in animals (Roberts et al., 1992Go, 1994Go), raise the possibility that ED shifting and reversal learning depend on dissociable neurochemical substrates, with the former being dopamine-dependent and the latter dopamine-independent. Conversely, Cools et al. (2001Go) have demonstrated that levodopa impairs reversal learning whilst improving task-switching performance, suggesting that dopamine influences both cognitive processes in PD, although in opposing directions. However, it should be borne in mind that the tasks employed by Cools et al. differed from ours in several respects, and in particular reversal learning was examined using probabilistic rather than total feedback, thus their results cannot be directly compared with ours.
A further important aspect of the current study is its capacity to examine the interaction between COMT genotype, an endogenous determinant of cortical synaptic dopamine levels, and exogenous dopaminergic medication. An increase in levodopa dose was associated with deterioration in certain aspects of performance, as anticipated, but also a complex interaction between medication dose and COMT genotype was clearly demonstrable.
Specifically, in those with high activity COMT genotypes (val/val) and hence lower endogenous cortical dopamine levels, exogenous dopamine had a detrimental effect on performance through increasing the number of errors whilst searching for the target, whereas in those with higher endogenous cortical dopamine levels (met/met), additional exogenous dopaminergic stimulation had no effect (Fig. 3). The imaging data essentially mirrored these behavioural results: higher levodopa doses had a greater negative impact on activation in val compared to met homozygotes (Fig. 6B).
These data can be accommodated in the well-established hypothesis of an inverted U-shaped relationship between dopamine levels and prefrontal function (Fig. 7). The existence of such a relationship is supported by a wealth of evidence from both in vivo studies demonstrating non-linear effects of the iontophoretic application of D1 agonists as well as from behavioural and functional imaging studies in humans with genetically determined differences in prefrontal dopamine (reviewed in Williams and Castner, 2006Go). In healthy controls and patients with schizophrenia, an increasing number of met alleles is associated with improved prefrontal function (Egan et al., 2001Go; Malhotra et al., 2002Go; Blasi et al., 2005Go), as predicted by the left-hand side of the inverted U-shaped curve. In contrast, our findings suggest a detrimental effect of elevated cortical dopamine levels on prefrontal function, thus supporting our previous suggestion that patients with early PD are on the right-hand side of this curve (Foltynie et al., 2004bGo), consistent with the demonstration of a hyperdopaminergic state in the PFC in early PD (Rakshi et al., 1999Go; Kaasinen et al., 2001Go).
The ‘inverted U’ model further suggests that the effect of dopaminergic medication will differ according to an individual's pre-existing position on the curve. Indeed, we observed a greater detrimental effect of exogenous dopamine in PD val homozygotes who are expected to be nearer the peak of the curve than in PD met homozygotes who are already near the base of the curve, suggesting a floor effect in the latter group (Fig. 7).
Fig. 7 Hypothesized inverted U-shaped relationship between prefrontal function and dopamine levels. Position on the curve is likely to be determined by factors influencing baseline prefrontal dopamine level including pathology [e.g. PD, schizophrenia (SZ)] and COMT genotype. Our data support the suggestion that patients with early PD operate on the right-hand arm of the curve, with those homozygous for val nearer the peak and those homozygous for met nearer the base. Hence exogenous levodopa might be expected to produce a greater negative effect on prefrontal function in val/val than met/met PD patients.
The underlying neural mechanism of the inverse relationship between dopamine levels and executive performance in early PD is uncertain, although a ‘dopamine overload hypothesis’ has been proposed. Specifically, in the presence of high ‘tonic’ dopamine levels in the PFC in early disease, further elevation in prefrontal dopamine may have detrimental functional consequences due to a down-regulation of the neural response to ‘phasic’ dopamine, i.e. a reduced SNR (Grace, 1993Go). It has been proposed that tonic dopaminergic transmission reflects the activation of extrasynaptic D1 receptors, whilst post-synaptic D2 activation mediates phasic dopamine signalling (Cohen et al., 2002Go), thus a reduced SNR in the ‘dopamine overload’ state may reflect a supra-optimal D1/D2 activation ratio. As the disease advances, there is a reduction in prefrontal dopamine storage (Brooks and Piccini, 2006Go) presumably due to the loss of mesocortical dopaminergic projections.
The ‘dopamine overload’ effect is therefore likely to disappear, such that increases in prefrontal dopamine and the D1/D2 activation ratio are beneficial rather than detrimental in terms of executive performance.
Irrespective of the underlying neural mechanisms, it seems that the effects of dopamine on cognitive performance must be interpreted in light of an individual's pre-existing position on a U-shaped curve, which is likely to be influenced by pathology within the dopaminergic system (such as schizophrenia or PD) as well as COMT genotype.
Dopaminergic activity in the PFC also declines with advancing age (Kaasinen and Rinne, 2002Go) and with disease duration in PD (Brooks and Piccini, 2006Go), further complicating the issue. In addition, it has been suggested that the optimal range of dopamine signalling (i.e. position of the peak of the curve) varies as a function of the nature of the task in question (Williams and Castner, 2006Go), which would be in keeping with the reported observation that the COMT polymorphism has differential effects according to task demand. Specifically, Nolan and colleagues have demonstrated that in healthy controls, met alleles improve performance on tasks requiring cognitive stability, whilst impairing performance on tasks dependent on cognitive flexibility (Nolan et al., 2004Go).
As might be expected, we observed an opposite pattern in our PD patients, with met alleles impairing ability to form an attentional set (reflecting cognitive stability) whilst having a tendency to improve ability to shift from one stimulus dimension to another (reflecting cognitive flexibility; see Fig. 2).
We have speculated that the underactivation that we observed within the frontoparietal network in PD met homozygotes reflects impaired prefrontal function. Other studies have similarly reported a decrease in prefrontal BOLD response in association with impaired executive performance in both schizophrenics (Callicott and Weinberger, 1999Go) and patients with PD (Lewis et al., 2003bGo; Williams-Gray et al., 2007bGo). However, some authors have observed reductions in BOLD response in association with stable or improved behavioural performance on working memory tasks, and thus argue that such changes might reflect an increase in cortical efficiency (Egan et al., 2001Go; Mattay et al., 2003Go; Blasi et al., 2005Go). Callicott and colleagues attempted to specifically investigate this conundrum by exploring fMRI activation in controls and schizophrenic patients divided into subgroups according to performance ability on the N-back working memory task. Their data suggest that prefrontal circuitry is under-recruited in patients whose task performance is poor, whereas those that perform well alter their neural recruitment pattern through compensatory overactivation of other prefrontal areas (Callicott et al., 2003Go). In a further recent study employing an attentional shifting task, Monchi and colleagues reported areas of both increased and decreased PFC activation in PD patients versus controls in association with impaired performance. Decreased activation was reported during those components of the task requiring caudate activation, whereas increased activation was reported when the caudate was not involved. The authors suggest that directionality of the change in BOLD response might reflect whether the cognitive process is dependent on corticostriatal or mesocortical networks (Monchi et al., 2007Go). Hence the relationship between BOLD activation and behavioural performance might vary according to both behavioural ability of the subjects and task demand.
In this study, we have focused principally on the effects of COMT genotype and dopaminergic medication on frontoparietal activity, given that firstly, the task generated significant BOLD activation exclusively within this cortical network, and secondly the dopamine-regulating influence of COMT appears to be confined to cortical areas (Karoum et al., 1994Go; Gogos et al., 1998Go; Mazei et al., 2002Go) due to low numbers of dopamine transporters there (Lewis et al., 2001Go). However, it is possible that COMT also exerts some of its effects on set-shifting performance in PD through subcortical mechanisms, despite our observation that genotype had no effect on caudate activation during the task. Recent evidence from a multimodal imaging study suggests that COMT genotype influences the interaction between prefrontal activation during a working memory task and midbrain dopamine synthesis, with a negative correlation in met homozygotes but a positive correlation in val carriers. Hence COMT may be involved in a mesocortical tuning mechanism which aims to maintain prefrontal dopamine at an optimum level for cognitive performance (Meyer-Lindenberg et al., 2005Go).
A final point for consideration is the use of equivalent levodopa dose as a quantitative measure of total dopaminergic medication. This measure is open to criticism as it may not accurately reflect frontal dopaminergic stimulation in individuals with different COMT genotypes, given that levodopa is subject to the metabolic effects of COMT whereas dopamine agonists are not. However, a recent study exploring the impact of the COMT polymorphism on levodopa pharmacokinetic-pharmacodynamic response patterns failed to identify any differences between genetic subgroups (Contin et al., 2005Go). Furthermore, it is clearly necessary to account quantitatively for the impact of dopamine agonists in some way, thus we feel that the calculated equivalent levodopa dose provides the best achievable estimate of overall dopaminergic stimulation.
In summary, we have demonstrated that a single functional polymorphism within the COMT gene alters the strategy adopted during an attentional shifting task in PD patients through altering activation in frontoparietal networks. Furthermore, the effect of dopaminergic medication on both performance and cortical activation is influenced by COMT genotype, as predicted by the inverted U hypothesis relating dopamine levels to prefrontal function. This work therefore reveals for the first time that attentional control in PD is crucially dependent on genetically determined and pharmacological variations in dopamine within frontoparietal networks, and clarifies the reasons underlying heterogeneity of executive function in PD. Furthermore, this study highlights the risk of medication-induced cognitive dysfunction in certain genotypic groups of patients, which may ultimately have implications for clinical practice.